NL2032338B1 - Intelligent decision-making system for pavement maintenance and repair - Google Patents
Intelligent decision-making system for pavement maintenance and repair Download PDFInfo
- Publication number
- NL2032338B1 NL2032338B1 NL2032338A NL2032338A NL2032338B1 NL 2032338 B1 NL2032338 B1 NL 2032338B1 NL 2032338 A NL2032338 A NL 2032338A NL 2032338 A NL2032338 A NL 2032338A NL 2032338 B1 NL2032338 B1 NL 2032338B1
- Authority
- NL
- Netherlands
- Prior art keywords
- repair
- maintenance
- road
- pavement
- schedule
- Prior art date
Links
- 238000012423 maintenance Methods 0.000 title claims abstract description 265
- 230000008439 repair process Effects 0.000 title claims abstract description 212
- 238000004458 analytical method Methods 0.000 claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 10
- 230000002787 reinforcement Effects 0.000 claims abstract description 9
- 239000000463 material Substances 0.000 claims description 18
- 230000003449 preventive effect Effects 0.000 claims description 16
- 238000003801 milling Methods 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 8
- 238000007689 inspection Methods 0.000 claims description 5
- 238000000556 factor analysis Methods 0.000 claims description 3
- 230000003746 surface roughness Effects 0.000 claims 1
- 238000004381 surface treatment Methods 0.000 claims 1
- 230000008901 benefit Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 10
- 230000007480 spreading Effects 0.000 description 9
- 230000000875 corresponding effect Effects 0.000 description 8
- 230000009471 action Effects 0.000 description 7
- 239000010410 layer Substances 0.000 description 6
- 238000007789 sealing Methods 0.000 description 6
- 230000008569 process Effects 0.000 description 5
- 238000007726 management method Methods 0.000 description 4
- 238000005056 compaction Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 3
- 238000005096 rolling process Methods 0.000 description 3
- 239000004576 sand Substances 0.000 description 3
- 239000010426 asphalt Substances 0.000 description 2
- 239000003292 glue Substances 0.000 description 2
- 239000002344 surface layer Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- VXPSARQTYDZXAO-CCHMMTNSSA-N (4s,4ar,5s,5ar,12ar)-4-(dimethylamino)-1,5,10,11,12a-pentahydroxy-6-methylidene-3,12-dioxo-4,4a,5,5a-tetrahydrotetracene-2-carboxamide;hydron;chloride Chemical compound Cl.C=C1C2=CC=CC(O)=C2C(O)=C2[C@@H]1[C@H](O)[C@H]1[C@H](N(C)C)C(=O)C(C(N)=O)=C(O)[C@@]1(O)C2=O VXPSARQTYDZXAO-CCHMMTNSSA-N 0.000 description 1
- 241000282326 Felis catus Species 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 239000003795 chemical substances by application Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000007728 cost analysis Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000009429 distress Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000013439 planning Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000004224 protection Effects 0.000 description 1
- 239000000523 sample Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Road Repair (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses an intelligent decision—making system for pavement maintenance and repair. The system includes a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair scheme generation module, a road maintenance and repair scheme analysis module and a road maintenance optimal scheme output module. According to the present invention, expert decision—making knowledge and a reinforcement learning algorithm are combined, a maintenance decision—making method is continuously optimized, correctness of maintenance measures is ensured, and maintenance benefits are improved.
Description
P1444 /NLpd
INTELLIGENT DECISION-MAKING SYSTEM FOR PAVEMENT MAINTENANCE AND
REPAIR
The present invention relates to the field of road mainte- nance, and in particular to an intelligent decision-making system for pavement maintenance and repair.
In practical work, highway managers do a lot of testing work and collect a large amount of testing data during daily mainte- nance and special maintenance. However, many valuable data have not been effectively used and are only used for simple statistics, which cannot be correlated with other multi-source maintenance da- ta and are subjected to in-depth mining and analysis. Therefore, the causes of pavement distress and the law of space-time expan- sion cannot be grasped, leading to weak data support for the deci- sion-making of maintenance and repair schemes. In addition, when many maintenance and repair measures can be selected in the face of pavement damage, which to select cannot be determined because of the lack of scientific selection basis, so that a maximum maintenance benefit of pavement cannot be achieved.
An objective of the present invention is to provide an intel- ligent decision-making system for pavement maintenance and repair.
The system includes a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair scheme generation module, a road mainte- nance and repair scheme analysis module, and a road maintenance optimal scheme output module.
The road state parameter acquisition module acquires state parameters of a road to be maintained and repaired and sends the state parameters to the road maintenance and repair scheme genera- tion module.
The state parameters of the road to be maintained and re- paired include a road age, a road type, a pavement structure type, surface composition, a traffic condition, a road grade, regional division, a damage rate (DR)/a pavement condition index (PCI), a pavement average international roughness index (IRI)/a riding quality index (RQI), a pavement structure representative deflec- tion (DEF) /a pavement structure strength index (PSSI), an average maximum rutting depth (RD)/a rutting depth index (RDI), a pavement skid resistance index (SRI), a first main damage type and a second main damage type.
The maintenance and repair reference factor acquisition mod- ule acquires a maintenance and repair reference factor and sends the maintenance and repair reference factor to the road mainte- nance and repair scheme analysis module.
The maintenance and repair reference factor is determined by expert knowledge, and clustering factor analysis is conducted to determine an effective category field. A scoring standard of the maintenance and repair reference factor is determined by the ex- pert knowledge and an industrial standard.
Preferably, the maintenance and repair reference factors in- clude a material used for implementing the road maintenance and repair scheme and a corresponding priority, as well as a grade, cost, a city, and a traffic grade of the road section to be main- tained and repaired.
The road maintenance and repair scheme generation module stores a road maintenance and repair scheme generation model.
The road maintenance and repair scheme generation model in- cludes a quadruple <S, A, P, R>. S indicates the state parameters of the road to be maintained and repaired. A indicates several road maintenance and repair reference schemes. P indicates a prob- ability of state change of a maintenance and repair position of pavement. R indicates a reward value function.
The road maintenance and repair reference scheme includes a repair scheme, a daily maintenance scheme, and a preventive maintenance scheme.
A daily maintenance scheme M1={daily inspection M1-1; daily maintenance M1-2; daily repair M1-3}; a preventive maintenance scheme M2= {sealing M2-1; functional overlay; M2-2; preventive maintenance combination M2-3}; and a repair scheme M3={milling and covering M3-1; structural reinforcement M3-2; surface overhaul M3- 3; base overhaul M3-4; subgrade treatment M3-5}.
The reward value function is indicated as:
Reg = WIRD + + wR 4 ow, REE (1)
In the formula, Ww; is a weight coefficient of a reward value of an ith performance index; Xi, Ww; =1, index_iindicates the ith per- formance index; i=1, 2, .., n; Rt41is a total reward value after a certain road maintenance and repair reference scheme is implement- ed; Ringer is a reward value of the ith performance index after a road maintenance and repair reference scheme is implemented; and a performance index is one or more of the state parameters.
Preferably, the reward value function is indicated as:
Ri41 = wpe REE + Wro REG +wap REPL. (2)
In the formula, Wpe, Wrgi and Wgp, are weight coefficients of reward values of PCI, RQI, and RDI.
The reward value RÎ of the PCI satisfies:
REEL = ¢1DRy 41 + C2Spr main + CsFor type (3)
In the formula, €. €. C3 are a pavement damage quantity coef- ficient, a pavement damage distribution coefficient, and a pave- ment damage main type coefficient; DR, is a pavement damage quan- tity; Sprmainis a pavement damage distribution, and Fpgtype is a pavement damage main type.
The reward value RX of the RQI satisfies:
RE = dyIRI, 1 + dyIRI max. (4)
In the formula, d,. d, are a pavement roughness condition co- efficient and an extreme roughness coefficient; IRh41is a pavement roughness condition; and IRI_max is an optimal pavement roughness.
The reward value REP! of the RDI satisfies:
RRP! = e,RD;,1 + e2RD_max. (5)
In the formula, €. €32 are a rutting coefficient and an ex- treme rutting depth coefficient; RD; is a rutting depth; and
RD_max is a maximum rutting depth.
The road maintenance and repair scheme generation model re- ceives the state parameters of the road to be maintained and re- paired and computes reward values under different road maintenance and repair reference schemes.
The road maintenance and repair scheme generation module writes a road maintenance and repair reference scheme having a re- ward value larger than a preset threshold into a road maintenance and repair scheme set and sends the road maintenance and repair reference scheme to the road maintenance and repair scheme analy- sis module.
The road maintenance and repair scheme analysis module evalu- ates a road maintenance and repair scheme in the road maintenance and repair scheme set according to the maintenance and repair ref- erence factor, determines a priority of the road maintenance and repair scheme, and sends the priority to the road maintenance op- timal scheme output module.
A method of evaluating the road maintenance and repair scheme in the road maintenance and repair scheme set by the road mainte- nance and repair scheme analysis module includes: using an actor- critic model to compute a reward (Rt+1) for each road maintenance and repair scheme, and sorting road maintenance and repair schemes in descending order according to a reward value Ru, so as to de- termine the priority of the road maintenance and repair scheme.
According to the present invention, a technical effect is un- doubted, expert decision-making knowledge and a reinforcement learning algorithm are combined, a maintenance decision-making method is continuously optimized, the correctness of maintenance measures is ensured, and maintenance benefits are improved.
FIG. 1 shows a flow of decision-making of a system.
The present invention is further described below in conjunc- tion with the embodiments, but it should not be understood that the scope of the subject of the present invention is limited to the following embodiments. Without departing from the technical idea of the present invention, various substitutions and changes may be made according to the common technical knowledge and con- ventional means in the art, which should be included in the pro- tection scope of the present invention. 5 Embodiment 1:
With reference to FIG. 1, an intelligent decision-making sys- tem for pavement maintenance and repair includes a road state pa- rameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair scheme generation module, a road maintenance and repair scheme analysis module, a road maintenance optimal scheme output module and a da- tabase.
The road state parameter acquisition module acquires state parameters of a road to be maintained and repaired and sends the state parameters to the road maintenance and repair scheme genera- tion module.
The state parameters of the road to be maintained and re- paired include a road age, a road type, a pavement structure type, surface composition, a traffic condition, a road grade, regional division, a damage rate (DR)/a pavement condition index (PCI), a pavement average international roughness index (IRI)/a riding quality index (RQI), a pavement structure representative deflec- tion (DEF) /a pavement structure strength index (PSSI), an average maximum rutting depth (RD)/a rutting depth index (RDI), a pavement skid resistance index (SRI), a first main damage type and a second main damage type (for example, the first main damage type is a transverse crack, and the second main damage type is a pit slot).
The maintenance and repair reference factor acquisition mod- ule acquires a maintenance and repair reference factor and sends the maintenance and repair reference factor to the road mainte- nance and repair scheme analysis module.
The maintenance and repair reference factor is determined by expert knowledge, and clustering factor analysis is conducted to determine an effective category field. A scoring standard of the maintenance and repair reference factor is determined by the ex- pert knowledge and an industrial standard.
In the embodiment, the maintenance and repair reference fac-
tors determined by the expert knowledge include a material used for implementing the road maintenance and repair scheme and a cor- responding priority, as well as a grade, cost, a city, and a traf- fic grade of the road to be maintained and repaired.
The road maintenance and repair scheme generation module stores a road maintenance and repair scheme generation model.
The road maintenance and repair scheme generation model in- cludes a quadruple <S, A, P, R>. S indicates the state parameters of the road to be maintained and repaired. A indicates several road maintenance and repair reference schemes. P indicates a prob- ability of state change of a maintenance and repair position of pavement. R indicates a reward value function.
The road maintenance and repair reference scheme includes a repair scheme, a daily maintenance scheme, and a preventive maintenance scheme.
A preventive maintenance scheme M1={daily inspection M1-1; daily maintenance M1-2; daily repair M1-3}; a daily maintenance scheme M2={sealing M2-1; functional covering M2-2; preventive maintenance combination M2-3}; and a repair scheme M3={milling and covering M3-1; structural reinforcement M3-2; surface overhaul M3- 3; base overhaul M3-4; subgrade treatment M3-5}.
The reward value function is indicated as:
Rey = Wei REF] + wrgrRÍS + WrorRELL (2)
In the formula, Wpcr: Wro: and Wpp; are weight coefficients of reward values of PCI, RQI and RDI.
The reward value of the PCI satisfies:
RP] = C1DRt+1 + C2Spr main + C3Fpr type - (3)
In the formula, €, Cz, ¢3 are a pavement damage quantity coef- ficient, a pavement damage distribution coefficient and a pavement damage main type coefficient; DR:1is a pavement damage quantity;
Spr mainls a pavement damage distribution; and Fpgtype 1s a pavement damage main type.
The reward value REY of the RQI satisfies:
REY = dyIRI yy + dyIRI max. (4)
In the formula, di. d, are a pavement roughness condition co-
efficient and an extreme roughness coefficient; IRlh41is a pavement roughness condition; and JRI max is optimal pavement roughness.
The reward value RED! of the RDI satisfies:
REP! = e,RD;41 + e2RD_max. (5)
In the formula, €. e; are a rutting coefficient and an ex- treme rutting depth coefficient; RD is a rutting depth; and
RD_max is a maximum rutting depth.
The road maintenance and repair scheme generation model re- ceives the state parameters of the road to be maintained and re- paired and computes reward values under different road maintenance and repair reference schemes.
The road maintenance and repair scheme generation module writes a road maintenance and repair reference scheme having a re- ward value larger than a preset threshold into a road maintenance and repair scheme set and sends the road maintenance and repair reference scheme to the road maintenance and repair scheme analy- sis module.
The road maintenance and repair scheme analysis module evalu- ates a road maintenance and repair scheme in the road maintenance and repair scheme set according to the maintenance and repair ref- erence factor, determines a priority of the road maintenance and repair scheme, and sends the priority to the road maintenance op- timal scheme output module.
A method of evaluating the road maintenance and repair scheme in the road maintenance and repair scheme set by the road mainte- nance and repair scheme analysis module includes the following steps that an actor-critic model is used to compute a reward (Rt+1) for each road maintenance and repair scheme, and road maintenance and repair schemes are sorted in descending order ac- cording to a reward value R‚1; so as to determine the priority of the road maintenance and repair scheme. Specifically, a state is obtained for a strategy, an action (At) is output, and then a new state (St+l) and the reward (Rt+1) are received. A critic model computes a score q of an action in a current state, and an actor model uses q to update an own strategic weight.
The road maintenance optimal scheme output module outputs the road maintenance and repair scheme having a highest priority as a road maintenance optimal scheme.
The database stores data from the road state parameter acqui- sition module, the maintenance and repair reference factor acqui- sition module, the road maintenance and repair scheme generation module, the road maintenance and repair scheme analysis module, and the road maintenance optimal scheme output module.
Embodiment 2:
An intelligent decision-making system for pavement mainte- nance and repair includes a road state parameter acquisition mod- ule, a maintenance and repair reference factor acquisition module, a road maintenance and repair scheme generation module, a road maintenance and repair scheme analysis module, and a road mainte- nance optimal scheme output module.
The road state parameter acquisition module acquires state parameters of a road to be maintained and repaired and sends the state parameters to the road maintenance and repair scheme genera- tion module.
The state parameters of the road to be maintained and re- paired include a road age, a road type, a pavement structure type, surface composition, a traffic condition, a road grade, regional division, DR/PCI, pavement average IRI/RQI, pavement structure representative DEF/PSSI, an average maximum RD/RDI, pavement SRI, a first main damage type and a second main damage type.
The maintenance and repair reference factor acquisition mod- ule acquires a maintenance and repair reference factor and sends the maintenance and repair reference factor to the road mainte- nance and repair scheme analysis module.
The maintenance and repair reference factors include a mate- rial used for implementing the road maintenance and repair scheme and a corresponding priority, as well as a grade, cost, a city, and a traffic grade of the road to be maintained and repaired.
The road maintenance and repair scheme generation module stores a road maintenance and repair scheme generation model.
The road maintenance and repair scheme generation model in- cludes a quadruple <S, A, P, R>. S indicates the state parameters of the road to be maintained and repaired. A indicates several road maintenance and repair reference schemes. P indicates a prob- ability of state change of a maintenance and repair position of pavement. R indicates a reward value function.
The road maintenance and repair reference scheme includes a repair scheme, a daily maintenance scheme and a preventive mainte- nance scheme.
A daily maintenance scheme M1l={daily inspection M1-1; daily maintenance M1-2; daily repair M1-3}; a preventive maintenance scheme M2= {sealing M2-1; functional covering M2-2; preventive maintenance combination M2-3}; and a repair scheme M3={milling and covering M3-1; structural reinforcement M3-2; surface overhaul M3- 3; base overhaul M3-4; subgrade treatment M3-5}.
The reward value function is indicated as:
Rei = wR 4 RI 4 wy RIMMER (1)
In the formula, w; is a weight coefficient of a reward value of an ith performance index; NW; =1, index_iindicates the ith per- formance index; i=1, 2, .., n; Ris a total reward value after a road maintenance and repair reference scheme is implemented; Rindex.i is a reward value of the ith performance index after a road maintenance and repair reference scheme is implemented; and a per- formance index is one or more of the state parameters.
The road maintenance and repair scheme generation model re- ceives the state parameters of the road to be maintained and re- paired and computes reward values under different road maintenance and repair reference schemes.
The road maintenance and repair scheme generation module writes a road maintenance and repair reference scheme having a re- ward value larger than a preset threshold into a road maintenance and repair scheme set and sends the road maintenance and repair reference scheme to the road maintenance and repair scheme analy- sis module.
The road maintenance and repair scheme analysis module evalu- ates a road maintenance and repair scheme in the road maintenance and repair scheme set according to the maintenance and repair ref- erence factor, determines a priority of the road maintenance and repair scheme, and sends the priority to the road maintenance op-
timal scheme output module.
The step of evaluating a road maintenance and repair scheme in the road maintenance and repair scheme set by the road mainte- nance and repair scheme analysis module includes the following steps that for maintenance scheme decision-making, historical behaviors (that is, specific maintenance measures) are recorded as A; and performance P of gradual decay over time is observed, initial re- wards of all decision-making schemes are 0, the maintenance and repair measures that are determined to be effective obtain corre- sponding reward values, and the reward values are added.
The road maintenance optimal scheme output module outputs the road maintenance and repair scheme having a highest priority as a road maintenance optimal scheme.
Embodiment 3:
An intelligent decision-making system for pavement mainte- nance and repair uses a two-level decision-making strategy, and a specific solution process is shown in the figure. First-level de- cision-making is scheme-level decision-making, and mainly serves network-level facility management, such as maintenance planning, and local and regional management; and second-level decision- making is measure-level decision-making, and mainly serves pro- jJect-level facility management, such as decision-making assistance of a maintenance department or a maintenance design unit.
The scheme-level decision-making is mainly based on actual attributes and state of a road, takes scientific decision-making as a target, and determines a decision-making conclusion most suitable for a current road section; and the measure-level deci- sion-making includes selection of measures and materials and cost analysis, mainly considers regional features, an environment, traffic conditions, a capital restriction and a maintenance level of the road, and uses user habits to indicate decision-making hab- its and advantages of materials and devices of local, regions and enterprises, and other potential information.
A road maintenance and repair decision-maker determines which project level of maintenance and repair to conduct by means of a combination state of boundary values of macro indexes. The project levels include an overhaul project, a repair project, a preventive maintenance project, and daily maintenance and repair.
The decision-making system creates four levels of maintenance and repair measure sets, and each level corresponds to a set of a sublevel of the level. A first level is called a project level; a second level includes maintenance and repair schemes of different degrees, and is called a scheme level; a third level includes dif- ferent maintenance and repair measures, and is called a measure level; and a fourth level includes optional material combinations of different measures, and becomes a measure material level. A set relation between the project level and a scheme is as follows:
A partial repair and overhaul project (a repair project) M3= {milling and covering M3-1; structural reinforcement M3-2; surface overhaul M3-3; base over- haul M3-4; subgrade treatment M3-5}; a preventive maintenance project M2= {sealing M2-1; function- al covering M2-2; preventive maintenance combination M2-3}; and daily maintenance Ml={daily inspection M1-1; daily mainte- nance M1-2; daily repair M1-3}.
There is a strong correlation between scheme-level measures and a current road structure situation, so a primary decision- making problem of the project is scheme decision-making, that is, a detailed diagnosis of a road structure and a current damage sit- uation. 1) Markov decision process (MDP) model-scheme decision-making
The MDP model consists of a quadruple <S, A, P, R>. S indi- cates a state and represents various current indexes and parame- ters of pavement; A indicates an action and represents different maintenance and repair schemes; P indicates a probability of state change and represents the probability of the state change of a maintenance and repair position of the pavement; and R indicates a reward value function and represents rewards and punishments ob- tained by an intelligent body after conducting actions.
For each index, according to data actually stored, primary detection index data is preferred, so as to prevent possible er-
rors caused by computation and transformation of evaluation index- es and a non-universal index problem caused by change of an indus- try evaluation standard. Indexes related to damage are mainly used for correlation analysis of road damage and dimensionally reduced.
Rutting damage is basically independent of other damage types, so it is illogical to classify damage simply according to defor- mation, damage and surface features.
Therefore, the system determines that various parameters of the MDP model related to the damage include:
S={a road age; a road type; a pavement structure type; sur- face composition; a traffic condition; a road grade; regional di- vision; DR/PCI; pavement average IRI/RQI; pavement structure rep- resentative DEF/PSSI; average maximum RD/RDI; pavement SRI; a first main damage type; a second main damage type}, and feature indexes of the 14 dimensions are mainly included.
A={M3-1; M3-2; M3-3; M3-4; M3-5; M2-1; M2-2; M2-3;
M1-1; M1-2; M1-3}
According to decay of main dynamic indexes of a road after the road uses a certain maintenance and repair scheme, a punish- ment or reward is given, and target service lives and performance decay thresholds are set for different maintenance and repair schemes. A reward is given when pavement performance is improved and a performance decay rate or threshold does not exceed a set range; and a punishment is given when performance improvement measures fail quickly and a threshold is exceeded. The reward val- ue function is defined as:
Res = wy RIX pw RIMdext yw, Rex, where w; is a weight coefficient of a reward value of each index; index i is an ith performance index, there are n indexes in total, and X.,w;=1; and an index for a reward may be selected ac- cording to an actual situation of an actual road tracking index.
The system selects the PCI, RQI and RDI as long-term tracking in- dexes, and the reward value function is:
Resi = WperREE + WrorRE + wep RED, where RPE! indicates that a road pavement damage condition does not include serious decay, and includes partial damage, which is within a rational range. A specific definition is:
REE = Ci DRt11 + C25pR main + C3FpR typer where ¢;. €2, C3 are a pavement damage quantity coefficient, a pavement damage distribution coefficient and a pavement damage main type coefficient; and
Rie indicates that road pavement driving quality does not decay seriously, and a local position is uneven, which is within a rational range. A specific definition is:
RY = d;IRl41 + d2IRI_ max, where dy, d; are a pavement roughness condition coefficient and an extreme roughness coefficient; and
RRP! that a road pavement rut does not decay seriously, and a local position is uneven, which is within a rational range. A spe- cific definition is as follows:
REPI = e, RD, + e2RD_max, where ey. e, are a pavement roughness condition rutting coef- ficient and an extreme rutting depth coefficient. 2) Parameter dynamic solution
In maintenance strategy treatment, a repair project and pre- ventive maintenance may have corresponding project records, but within a daily maintenance project range, even if large-scale dai- ly repair is done, overall records are also lacking. Therefore, in data processing, a scheme is revised for daily maintenance facili- ties, data is checked according to a repair situation of surface damage, for a situation with a large change in a surface damage quantity, a daily repair strategy is conducted, daily maintenance is the default for high-grade roads (such as an expressway, a first-grade road, a second-grade road, an urban expressway, and an urban main road) without any maintenance strategy, and daily in- spection is default for low-grade roads without any maintenance strategy (other roads than the high-grade roads). In data execu- tion, it should also be noted that, in whole life cycle management of actual road operation and maintenance, even if there is no spe- cial maintenance and repair project, action A needs to be given at certain time intervals according to a maintenance level.
States are the road pavement dynamic and static parameters of
14 dimensions, an action is a used maintenance and repair scheme, and 12 1-dimensional options are maintenance and repair schemes actually used in history. In reinforcement learning, a fitted-Q learning algorithm is used to learn an optimal strategy. Because a proportion of daily maintenance projects is much higher than that of other strategies and data samples of the maintenance and repair schemes are few, in a process of randomly selecting learning sam- ples, groups are selected to solve parameters and use the parame- ters for all projects.
During inverse reinforcement learning, an initial parameter is set as 0, so as to eliminate influence of the initial parameter (prior knowledge}. All process reward parameters and weight values of are limited in [0,1]. In addition, some indexes select a weight parameters in an existing evaluation standard as prior knowledge to initialize the weight parameters as controls, which is used to increase a convergence rate. 3) Measure decision-making based on habits
In the same maintenance and repair scheme set, cost and effi- ciency differences between different measures are large, and there is no complete linear correlation between a maintenance and repair decision-making scheme and maintenance and repair cost. Therefore, selection of measures cannot be made solely according to a condi- tion of facilities, and factors such as economy and user habits have to be comprehensively considered. According to the learning process, in addition to parameters that may be determined such as related traffic, environment and maintenance level, a priority of measures is added to indicate the user habits.
A specific measure and material set and a field may also be dynamically adjusted according to a user situation. The exemplary measures table is shown in the following table. Actions that are taken involve three parameters: recommended measures, material combinations and a project unit price; and state parameters in- clude a user attribute (a capital source restriction), a road grade (high and low), a maintenance level (which is decided joint- ly by a road grade and a user attribute, is a non-independent pa- rameter and an implicit parameter, and does not need to be input), a city, a traffic grade and a user habit (priority assignment of measures may be reflected in the selection of material priority types, such as recycled and modified asphalt).
Table 1 Exemplary measures table I
Corresponding treatment measures Material type code code po foe ee [SET
M_ 13 Crack sealing M 132 glue {-20)
Severe cold crack sealing glue (-40}
Cold material cold repair Default
Material filling Functional covering material
Local milling and repaving Functional covering material en eenen [a ae
M 13 Asphalt surface layer layered repaving | M_1 3 15 layer
Scheme Measure
Corresponding treatment measures Material type code code
Local milling, interlayer removing and
M_ 13 M_1 3 19 | Default repaving
Local milling, base treatment and
M_1_3 M_1 3 20 | Default repaving
Local milling, base treatment or re-
M_ 13 M_1 3 21 | Default placement and repaving
Milling, digging, repaving and anti-
M_ 13 M_1 3 22 | Default stripping agent adding
Spreading of 3 mm-5 mm of grav-
M13 1-layer spreading M_1 3 23 el/coarse sand
Spreading of 5 mm-10 mm of grav- el, stable rolling compaction and
M_ 13 2-layer spreading M_1 3 24 spreading of 3 mm-5 mm of grav- el/coarse sand
Table 2 Exemplary measures table II
Scheme Measure
Corresponding treatment measures Material type code code
Spreading of 10 mm-15 mm of gravel, stable rolling compaction, spreading of 5 mm-10 mm of
M_13 3-layer spreading M_1 325 gravel, stable rolling compaction and spreading of 3 mm-5 mm of gravel/coarse sand
Paving of 1 cm-2 cm of microsur-
M 13 Local milling and paving M_ 1326 face
Paving of 1 cm-2 cm of ultrathin
M 13 Local milling and paving M_ 1 3_26 cover or thin cover
Local surface overhaul M_1 3 28 | Default
M_2 1 Sand-contained fog seal M_ 2 1 1 | Default nas Surface layer milling and covering-2.5 wos
M31 M313 cm sn ee — essen
M 32 M322 Default layer milling and 2-layers adding)
Claims (9)
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110843921.4A CN113723753A (en) | 2021-07-26 | 2021-07-26 | Intelligent decision-making system for pavement maintenance |
Publications (2)
Publication Number | Publication Date |
---|---|
NL2032338A NL2032338A (en) | 2023-01-31 |
NL2032338B1 true NL2032338B1 (en) | 2023-11-10 |
Family
ID=78674064
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
NL2032338A NL2032338B1 (en) | 2021-07-26 | 2022-06-30 | Intelligent decision-making system for pavement maintenance and repair |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN113723753A (en) |
NL (1) | NL2032338B1 (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114444737B (en) * | 2022-02-18 | 2024-05-28 | 招商局重庆交通科研设计院有限公司 | Pavement maintenance intelligent planning method based on transfer learning |
CN116797102A (en) * | 2023-07-21 | 2023-09-22 | 西安科技大学 | IRI-based pavement maintenance decision determining method and system |
-
2021
- 2021-07-26 CN CN202110843921.4A patent/CN113723753A/en active Pending
-
2022
- 2022-06-30 NL NL2032338A patent/NL2032338B1/en active
Also Published As
Publication number | Publication date |
---|---|
CN113723753A (en) | 2021-11-30 |
NL2032338A (en) | 2023-01-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
NL2032338B1 (en) | Intelligent decision-making system for pavement maintenance and repair | |
Madeh Piryonesi et al. | Using machine learning to examine impact of type of performance indicator on flexible pavement deterioration modeling | |
Wang et al. | Life cycle energy consumption and GHG emission from pavement rehabilitation with different rolling resistance | |
CN111062648B (en) | Evaluation method for comprehensive performance of asphalt pavement | |
CN113191660A (en) | Intelligent decision-making method for maintaining asphalt pavement of highway | |
Adel et al. | Developing parametric model for conceptual cost estimate of highway projects | |
Rebelo et al. | Use of data mining techniques to explain the primary factors influencing water sensitivity of asphalt mixtures | |
Bosurgi et al. | Optimizing artificial neural networks for the evaluation of asphalt pavement structural performance | |
Setianingsih et al. | Road maintenance and rehabilitation program using functional and structural assessment | |
Alnaqbi et al. | Predictive models for flexible pavement fatigue cracking based on machine learning | |
Ewadh et al. | Conflict to study safety at four leg-signalised intersections | |
Vasiliev et al. | Intellectualization of the monitoring process of street and road network objects | |
Papageorgiou et al. | Defining threshold values for pavement surface characteristics | |
Chassiakos et al. | Development of decision-support system for managing highway safety | |
Merritt et al. | Developing crash-modification factors for high-friction surface treatments | |
Paplauskas et al. | Road pavement condition index deterioration model for network-level analysis of national road network based on pavement condition scanning data | |
Sirvio | Advances in predictive maintenance planning of roads by empirical models | |
Fuentes | An analysis of sensitivity in economic forecasting for pavement management systems | |
Yichang et al. | Study of Georgia’s pavement deterioration/life and potential risks of delayed pavement resurfacing and rehabilitation. | |
Tidjani et al. | Impact of road infrastructure characteristics on road markings | |
Tadubana | Prediction of fuel consumption of haulage trucks in open pit mines | |
Musey | Quantifying the safety impact of high friction surface treatment installations in Pennsylvania | |
Uglova et al. | Reliability-Oriented Method of Flexible Pavement Management | |
Yuan et al. | Safety evaluation and improvements for highway intersections | |
Lazic | From Road Condition Data Collection to Effective Maintenance Decision Making |